Irish Times opinion piece by Colm Fagan: "Auto-enrolment plan seriously flawed"

[Duke of Marmalade]

Red herring from the man who brought in the tax treatment. Hmmm.

This is a great way to have an online debate alright. Attack any inconvenient comments as red herrings and take it off-line. Hmmm.

And one wonders why there's a lack of buy-in?!

I admitted that the tax angle was off topic and made my apologies.
Boss raised the possibility that a future greedy government would grab any "surplus". It was a valid, though not very well informed, on topic throwaway at the time. I have now pointed out that there never will be such a surplus and that in any case any discriminatory raid of the AE funds would be unconstitutional (provided they are set out legally as not being in any way part of the public sector).
Along the way a "debate" has developed as to whether the pension levy was a tax or a theft. I hope you agree that it is now off topic.

 

[Duke of Marmalade]

And one wonders why there's a lack of buy-in?!

I must address this separately. I suppose by positioning myself as a sort of Devil's Advocate against the simplistic knee jerk reactions (smoothing doesn't work, surpluses will be an easy smash and grab, Joe Duffy will have a field day etc.) it is understandable that you see me as a convinced proselyte of Colm's proposal. I see the practical objections though I have never seen a categorical demonstration that it doesn't work. So yes I am 100% behind Colm's request for a multi disciplinary review of his proposal by the likes of the ESRI.
There's a tad "shoot the messenger" tone to your last post which disappoints me given the more reasoned approach you demonstrated hitherto.

 

[Brendan Burgess]

Guys. this is a very important topic.

Can we please keep from attacking other posters. If you are attacked, don't respond in kind. Deal with the point and ignore the attack. (I know it's hard to do.)

Brendan

 

[Duke of Marmalade]

I think I have made my views clear. But @JimmyB99's post has alerted me to the fact that the way I express them is probably counterproductive. I have certainly said enough on this topic and I will leave it at that but will continue to follow the debate of course.

 

[Colm Fagan]

As promised, I'll try to get round to as many of the posts as possible. I must admit, though, that the quality of the contributions is below what I would have expected from AAM contributors based on my past experience. It is clear, sadly, that many of the posters have not actually read what I've written, as evidenced by the fact that no-one has answered my challenge from earlier this morning to tell me where they disagree with what I actually wrote as opposed to what they think I wrote.
Despite that, I'll soldier on - for now anyway. In future, I'll think twice before posting anything on AAM. The quality of discussion is far better on other forums. Maybe it's because they read what I actually wrote.
I don't know where to start. Let's try by looking at Table 1 of my paper, and try to use that to address some of the questions raised.

This table shows that, if the scheme started on 1 January 2020, with cash flows increasing from 10 in month 1 to 20 in month 2, 30 in month 3, etc. (probably not far off how the pattern of actual cash-flow roll-outs might look like), then monthly smoothed returns vary from a low of +0.13% to a high of +0.33%. Monthly market returns, by contrast, vary from a low of -15.1% to a high of +4.9%.

The problem is that a "smoothed" value of a fund is fine in theory but impossible to implement in practice. Everyone knows that financial markets fluctuate and are prone to prolonged optimism and pessimism. The problem is that where we are in the cycle is only clear in a decade's time and you have to make a drawdown tomorrow. I've been in a job where making policy depended on estimates of where we were the financial cycle and trust me it was really hard. Endless discussion about models and an impossibility of explaining them to people who made decisions.

Let's look at the above table in the light of @NoRegretsCoyote 's comments.
"Everyone knows that financial markets fluctuate". It's not often that they have fluctuated more than they did in the first few months of 2020, as the world came to grips with the financial impact of Covid19, yet the formula comes up trumps. "The problem is that where we are in the cycle is only clear in a decade's time and you have to make a drawdown tomorrow". When the smoothing formula was calculating the smoothed return for (say) March 2020, it didn't have a clue where in the cycle we were. I still don't know. It didn't have to know. All it required to know how to calculate any month's smoothed value were three things: 1. That month's market value; 2. The previous month's smoothed value; 3. The assumed long-term rate of return, which I've taken as 4% a year for all periods. It won't vary much from that in the long-term, unless there's a severe bout of inflation at some stage. The actual long-term return assumed is not a significant factor in determining returns, especially when the scheme is more mature.
"and you have to make a drawdown tomorrow" Let's assume that the net cash flow of 30 in March in the above table is 50 of gross inflows and 20 of gross outflows. The 20 who "made a drawdown tomorrow" got the smoothed value, which was significantly higher than market value. The excess over market value was paid for by the 50 gross who came in at that price. You can then ask: why were the 50 prepared to join at that inflated price? The answer is that, taking a long-term perspective, they could reasonably expect to earn smoothed returns in future like those shown above for the first six months of 2020. That gave them the confidence to stay in equities for life, including in the run-in to retirement and all through their retirement years. That peace of mind confers massive extra value and is well worth the cost of paying a bit over market value from time to time. On average, though, they know that, over their entire working lifetime and throughout their retirement, they will get - and receive - market value on average.
That table also helps to answer @nest egg 's comments about the Vanguard Balanced Fund.

If I had access in my pension to invest in say Vanguard's Global Balanced Fund, would I get more or less the same benefit as Colm's approach? A smoothed return.

I think that the term "balanced" in the fund's title refers to the fact that it has a mix of equities, bonds, and maybe some alternative assets. It does not refer to its performance. It would be interesting to compare its performance in the first six months of 2020 with the figures in the above table for smoothed and actual returns. I would be prepared to bet that its performance was very close to the unsmoothed returns e.g., a massive fall in value in March 2020. If you were invested in that fund and were due to retire at the end of March, you wouldn't feel very balanced. However, you would be more than happy if you were one of the 20 (netted from the 50 gross) under my smoothed approach who got out that month, having seen their fund grow by 0.13% in the month. Out of interest, @nest egg , could you get the actual numbers for me for the Vanguard Balanced Fund for those six months?

The second problem is that Colm's proposal seems to depend on pooling of current and future pensioners' funds.

I don't understand what the problem is here. The net cash flows of 10, 20, 30, 40, 50, 60 in months January to June 2020 in table 1 are the net result for active contributors (positive) and pensioners (negative). Of course, in that six months, the numbers of pensioners will be tiny, given that the fund only started in January, but you get the point. It doesn't matter a fig whether members are active or pensioners. Both groups are credited with exactly the same returns of +0.29% (Jan), +0.23% (Feb), etc.
OK, that's my lot for now. I may return to that same table to answer others' questions.

 

[Colm Fagan]

I would look at recreating it separately using derivatives on your smoothed index for a comparison.

Instead of the scheme, individual investors could buy and sell 'insurance' as options on the smoothed index with expiry of their specific retirement dates.

You buy a put option on the index being below the smoothed value on your retirement and sell a call option on it being above to help fund the put premium, it's a bit tricky because we don't know the smoothed value in advance, and as the market drops the smoothed value drops so e.g. the value of the put option drops. You'd need to price by simulation with an appropriate time series model on the market index.

Hi @RonnieShinbal88 . I get the general gist of what you're suggesting, but the detail is beyond my pay grade.
Your idea is probably fine in the context of a "normal" DC pension scheme, but AE is special. The restrictions on when and how much contributors can pay and when and how much claimants can withdraw make it very different from a normal scheme. My argument is that, with those restrictions in place, it is not necessary to go through the hoops you're suggesting to get a smoothed return. My simple approach does the trick, without ever having to go near derivatives. However, it only works for AE, not for anything else.
This brings me to a wider point, which I think will strike a chord.
When I was doing exams - a long time ago now - one of the key lessons I learned was the vital importance of reading the question carefully and using every piece of relevant information provided. If your solution didn't use all the relevant information, then your answer was wrong.
In considering how to design an auto-enrolment pension scheme, a key item of information in the question is the restrictions inherent in the scheme, in relation to both contributions and claims. Therefore, a solution that treats AE like an "ordinary" DC scheme must be wrong. The question demands a radically different solution.

 

[RonnieShinbal88]

Hi @RonnieShinbal88 . I get the general gist of what you're suggesting, but the detail is beyond my pay grade.
Your idea is probably fine in the context of a "normal" DC pension scheme, but AE is special. The restrictions on when and how much contributors can pay and when and how much claimants can withdraw make it very different from a normal scheme. My argument is that, with those restrictions in place, it is not necessary to go through the hoops you're suggesting to get a smoothed return. My simple approach does the trick, without ever having to go near derivatives. However, it only works for AE, not for anything else.
This brings me to a wider point, which I think will strike a chord.
When I was doing exams - a long time ago now - one of the key lessons I learned was the vital importance of reading the question carefully and using every piece of relevant information provided. If your solution didn't use all the relevant information, then your answer was wrong.
In considering how to design an auto-enrolment pension scheme, a key item of information in the question is the restrictions inherent in the scheme, in relation to both contributions and claims. Therefore, a solution that treats AE like an "ordinary" DC scheme must be wrong. The question demands a radically different solution.

I know almost nothing about AE schemes so probably shouldn't be commenting!

With derivatives I was just trying to put a price on what it would cost to recreate your scheme/ the smoothed index using the actual index plus derivatives, the cost would be non-zero due to people's risk aversion (and would pay the smoothed index value) but your scheme gets the same payoff for zero so it seems like a nice synergy (assuming an investor wants that payoff). I understand from your comment the difference now to DC and that you can't just insure one withdrawal date so it seems even better.

 

[NoRegretsCoyote]

it didn't have a clue where in the cycle we were. I still don't know. It didn't have to know.

The assumed long-term rate of return, which I've taken as 4% a year for all periods.

Colm these two statements are in complete contradiction.

The smoothing model is calibrated on long-run investment return assumptions. 4% is not God-given but based on observations over a long period. The market could return better or worse at which point you have to update your 4% assumption.

4% could even be too conservative and the actual value could spend 25 years above the smoothed value leaving a generation receiving a lower pension than if market value had been used. This would not be fair!

There is just no set of model parameters that is invariant to future market performance while always returning to equlibrium. There just isn't and you can't base a system to support the retirements of millions of workers pretending that there is.

 

[Colm Fagan]

Thanks @RonnieShinbal88
Your comment reminds of another driver for the smoothing proposal. I want to minimise seepage from various sources. Some we're familiar with: charges/ commissions by advisers (mainly on retirees), advisory fees on allocating funds between different asset classes, the cost of administering unit-linked funds (NEST in the UK has over 40 default funds, each having to be priced daily), excessive charges (mostly borne by retirees in the form of retail rather than group charges) for administration and asset management. Another source of seepage from member accounts is trading costs. My ideal for the fund (unattainable, I know) is a buy and hold strategy: it buys a share and holds it through thick and thin, possibly for 20, 30 years or longer. As Warren Buffett said: "My ideal holding period is forever" I'm sure your idea of puts and calls is very good, but the frictional cost of all that trading activity must be hefty, depleting members' funds. I haven't expressed it in this way, but one strand of my proposal could be seen as asking exiting and joining members to agree to swap assets at smoothed prices rather than at market prices, to their mutual long-term benefit. The rules of AE allow that pact to be enforced through thick and thin.

 

[Colm Fagan]

Hi @NoRegretsCoyote
I think you'll agree that a cycle typically spans four or five years, while an assumed long-term rate of return generally has a horizon of 20 to 40 years. In any event, if you go through the numbers in the example, you'll find that the assumed long-term return has a relatively small impact. In the projections, I assumed a long-term return of 4% a year for the entire projection period. It would have made little difference to the smoothed returns if I had assumed 5% instead. Come to think of it, I looked at exactly that difference in the paper, but I keep forgetting that people don't seem to want to read the paper.

 

[Colm Fagan]

4% could even be too conservative and the actual value could spend 25 years above the smoothed value leaving a generation receiving a lower pension than if market value had been used. This would not be fair!

I really wish you would take the time to read the paper. It demonstrates that the smoothed index crosses the market index many times over a short number of years, irrespective of whether markets are on long-term upward or downward trajectory. This is because, even in a bull market, there are occasional pull-backs, which are generally sufficient to bring smoothed and market value back into line. The same is true in reverse for a bear market.
The bear market effect can be seen clearly in the graph on page 13 for the Japanese market post 1990 (but you haven't looked at that, of course). A visual inspection indicates that the graph of smoothed returns crosses the graph of market returns around 7 times in the first 8 years. This is despite the market falling by around 60% in the period. The frequent cross-overs are helped by the fact that cash flows are being invested at lower prices as the market falls - another consequence of it being AE rather than a "normal" DC scheme.

 

[Colm Fagan]

@NoRegretsCoyote
Apologies for my snide comment in my last post about you not looking at the paper, and not noticing how the smoothed and market indices cross paths frequently. My explanation -not an excuse - is that I am frustrated, putting so much time and effort into writing the paper and responding to queries from posters, to find that they comment without having read the damn thing. I didn't aim the snide comment specifically at you. It was aimed more generally at critics who haven't studied the paper - and there are lots of them out there.
Maybe it's best for me not make any further comments. I just let myself get annoyed, which does nobody any good.
Goodbye everyone.

 

[Sarenco]

You can then ask: why were the 50 prepared to join at that inflated price? The answer is that, taking a long-term perspective, they could reasonably expect to earn smoothed returns in future like those shown above for the first six months of 2020.

Isn’t that a BIG assumption?

I would have thought it was far more likely that they would hold off contributing anything further until units are being offered at a discount again.

That would seem to be the logical approach at an individual level - no?

 

[Duke of Marmalade]

Isn’t that a BIG assumption?

I would have thought it was far more likely that they would hold off contributing anything further until units are being offered at a discount again.

That would seem to be the logical approach at an individual level - no?

Yes, and there would have to be rules against such gaming just as there are in health insurance. In any case they would be giving up the State and employer subsidies by suspending contributions. That could never be financially rational.

 

[NoRegretsCoyote]

if you go through the numbers in the example, you'll find that the assumed long-term return has a relatively small impact. In the projections, I assumed a long-term return of 4% a year for the entire projection period. It would have made little difference to the smoothed returns if I had assumed 5% instead.

It makes a huge difference! If you eyeball your own Figure 6 you see that use of a 4% assumption sees smoothed below actual values for nearly a decade-long stretch at the end. And at an earlier point the use of 4% shows a market value about 60% above smoothed value at some point.

You could recalibrate the value for p in the smoothing formula of course so that it adjusts much more rapidly to market values with smaller deviations. But then the smoothed value wouldn't be very smooth anymore!


@Colm Fagan in your paper you write:

in theory, the chosen value of “it” should be the trustees’ best estimate of the expected long-term return at time t, composed of the expected risk-free return plus the expected Equity Risk Premium (ERP). Whilst the risk-free component is readily observable in the market, the same is not true for the ERP. Even among experts, opinions vary widely32 . Therefore, careful governance will be required around its periodic assessment. The likelihood is that the trustees will reassess the expected long-term return only once a year, possibly less frequently.

So you are conceding that a committee of wise heads are going to be re-estimating the fundamental parameters of the formula as often as every 12 months. It's not clear to me whether this will be backward-looking or only prospective. Either way it completely undermines the notion that there is some kind of stable, theoretical smoothed value that can be estimated in real-time independent of market values.

There are lots of good ideas in your proposal but I just don't think your smoothing model can be deployed operationally.

 

[Colm Fagan]

It makes a huge difference! If you eyeball your own Figure 6 you see that use of a 4% assumption sees smoothed below actual values for nearly a decade-long stretch at the end. And at an earlier point the use of 4% shows a market value about 60% above smoothed value at some point.

Excellent question - and posed after studying the paper. Makes me break my self-imposed vow of silence - at least for now. It is also a good opportunity to explore an important aspect of the proposal.

As you know, what's important to the investor is not the absolute value of an index but the change in its value over time. The "5%" index in figure 8 (you wrote figure 6 but I think you meant figure 8) grows by 72.8% between years 15 and 20 - month 180 to month 240. That's an average growth rate of 11.56% a year. The "4%" index grows by 71.9% over the same period, an average of 11.44% a year. As you can see, the two growth rates are very close, implying that the choice of 4% or 5% assumed growth rate in the smoothing formula isn't important in the long run.
What you can't have, of course, is moving from the 4% index to the 5% index, i.e., moving from the blue line to the orange line in figure 8. That's what I meant by the statement that "careful governance will be required." The important point, as demonstrated by the graphs and my sample figures above, is consistency over time. This is an area where the economists in the ESRI will add value.

When you consider the consistency of smoothed returns, irrespective of the long-term growth assumptions, provided there is consistency over time, I hope that you'll reconsider your final comment about deploying my smoothing approach operationally.

 

[Colm Fagan]

The above exchange with @NoRegretsCoyote helps to illustrate another aspect of the smoothing proposal.
The particular five-year period I chose (purely by chance) for illustration purposes happened to be Dec 2004 to Dec 2009. As anyone who was around then will remember, that spans the Great Financial Crisis, when share prices fell through the floor. Yet under my proposal both smoothed graphs grew by an average of around 11.5% a year in that period. More importantly, there was hardly a ripple during the GFC. OK, the pace of growth reduced significantly, but returns were still positive. You'd be quite happy to see your 86-year old granny in that index, while you would have kept her as far away as possible from the unadjusted market index.
Yes, I know it can be difficult to get your head around all the implications of what I'm proposing, but I'm certain that the extra value IF we can take our courage in our hands and go for it, is AT LEAST 50% more than a more conventional approach to AE. Despite what some on this forum may think, I am not 100% certain that I'm right. However, given the scale of the gain for society and the state if I am right, it's well worth having the proposal evaluated in detail by the ESRI or a similar multi-disciplined economic/ financial/investment consultancy, to check if it holds water.
What must always be remembered, of course, is that this is only for auto-enrolment, and with the constraints as set out in section 2 of the paper to prevent the four types of anti-selection itemised in that section. It is not going to take over the world and wipe out financial advisers, pension consultants, life assurance companies, etc. They will still have plenty of business from non-AE schemes and individuals. Actually, they may even do better, because people will see more clearly the long-term advantages of equity investment.

 

[nest egg]

I think that the term "balanced" in the fund's title refers to the fact that it has a mix of equities, bonds, and maybe some alternative assets. It does not refer to its performance. It would be interesting to compare its performance in the first six months of 2020 with the figures in the above table for smoothed and actual returns. I would be prepared to bet that its performance was very close to the unsmoothed returns e.g., a massive fall in value in March 2020. If you were invested in that fund and were due to retire at the end of March, you wouldn't feel very balanced. However, you would be more than happy if you were one of the 20 (netted from the 50 gross) under my smoothed approach who got out that month, having seen their fund grow by 0.13% in the month. Out of interest, @nest egg , could you get the actual numbers for me for the Vanguard Balanced Fund for those six months?

You're correct, the title does refer to its composition (and that assets are automatically rebalanced to stay at specified allocations) and its performance should reflect that. The only data I can find is what I can gleam from Vanguard's site, its price fell 12% in March 2020, versus of fall of 18% for Vanguard's 100% equity fund. Some smoothing, but far less than in your approach.

Regarding getting out of the fund, perhaps I'm missing something but isn't a key tenant that you wouldn't face that situation, i.e. you should stay invested for the long term benefitting from its approach and economies of scale?

Last question, have you considered the potential for administrative errors, or even fraud in the calculation and/or communication of the formula/values to members? The governance and oversight would have to be water tight (@Brendan Burgess this may be another barrier as the mandarins may not want such responsibility)

 

[Colm Fagan]

Regarding getting out of the fund, perhaps I'm missing something but isn't a key tenant that you wouldn't face that situation, i.e. you should stay invested for the long term benefitting from its approach and economies of scale?

Last question, have you considered the potential for administrative errors, or even fraud in the calculation and/or communication of the formula/values to members? The governance and oversight would have to be water tight (@Brendan Burgess this may be another barrier as the mandarins may not want such responsibility)

Not sure I understand your first point. Everyone has to get out sometime. The way I look at it is that it's like a high-interest deposit account: money accumulates while you're working, it's withdrawn when you've retired - one big withdrawal at retirement date (the 25% tax-free cash) and the other 75% taken gradually over the rest of your days (between 3% and 8% a year, more than that allowed over age 80, plus optional longevity protection from age 75) . Of course, for the first 30 years or so of the scheme's existence, any monies withdrawn will be taken up by incoming (and continuing) contributors at prevailing smoothed values - the fund as a whole will be cash flow positive.
On your last question, I think the opposite will be the case to what you suggest, i.e., that there is far less potential for administrative errors with the smoothed approach.
With the smoothed scheme, there will be just one fund. Everyone, active and retired, will get exactly the same return each month (once a quarter should be sufficient, given the stability of 'interest rates'). Only three figures are needed each month to calculate that month's 'interest rate' and smoothed value: 1. current market value; 2. this month's cash flow and 3. last month's smoothed value. Everything flows from that.
Contrast that with the government's plans. They are proposing what they call a 'default fund', but in reality they'll need a whole range of default funds to cater for workers at different ages, i.e., funds with high equity content at young ages, funds with low equity content and high bond/cash content at older ages. The UK's NEST has over 40 different default funds. Ireland will need something similar. Everyone will also be able to opt for risk-rated funds (low risk, medium risk, high risk) irrespective of age. Each of these forty something funds will have to be valued every day and units allocated or de-allocated at the prices prevailing on those dates. That's a heck of a lot of calculations each day, which will have to be matched to records of each individual's stated preference (i.e., check that they have opted for a default or risk-rated fund, whether they've changed their options, etc.). It's a minefield administratively, with massive potential for mistakes, especially when you consider that most workers will be employed in small enterprises, many of which won't have efficient payroll systems and there won't be consistency in dates on which funds are transferred to asset managers relative to when the money was deducted from workers' and employers' accounts. Given the volatility of unit prices (see the table above showing volatility in the first few months of 2020, e.g. minus 15.1% in March, having been down 20% at one stage in the month), that could cause major problems. Contrast that with the smoothed proposal, where returns are stable from month to month (e.g. +0.13% in March 2020, +0.25% in April 2020, etc. ) so it doesn't matter much whether an individual contribution was credited to the member's account in say March or April.
The government's plans could also cause major problems of communications. Think, say, of two guys down in my local tyre repair shop, both of whom have chosen the default option but because they're different ages, they get different returns every month/day. That could cause confusion. With the smoothed approach, in contrast, every person in the entire country who's in the fund will get exactly the same return every month, irrespective of age, retirement status, value of their personal fund. They'll be chatting to one another regularly, in the pub, the hairdresser's, the coffee shop, about how disappointed they are when the "interest rate" falls to 0.13% in March 2020, having been 0.23% the previous month, but then they'll be heartened when they see the rate rising again in April 2020, to 0.25%. They will see those variations in return as significant, because that's their experience of seeing what happens to their savings in the local credit union or whatever. This type of person would be terrified out of their wits if someone told them that they'd lost over 15% of their money, as would have been the case if they had been invested in an equity unit-linked fund in March 2020.

 

[nest egg]

Not sure I understand your first point. Everyone has to get out sometime. The way I look at it is that it's like a high-interest deposit account: money accumulates while you're working, it's withdrawn when you've retired - one big withdrawal at retirement date (the 25% tax-free cash) and the other 75% taken gradually over the rest of your days (between 3% and 8% a year, more than that allowed over age 80, plus optional longevity protection from age 75) .

Crossed wires, I initially read your statement to mean that the member would be "getting out" of the smoothed fund. I now understand you mean moving from accumulation to drawdown phase, while staying invested in it.

As you've raised the tax free lump sum, if they don't need the money, wouldn't it be in the members' best interests not to take it? The temptation is obviously going to be great, and certainly a percentage of people may be banking on the sum to pay off a mortgage or help their kids, but inevitably some people will just end up moving the lump sum into a savings a/c, or into other investment funds (with all of the potential downsides you've pointed out), which would underperform the smoothed fund. Perhaps lump sums should be an "opt-in" to ensure a conscious decision is made by the member to make the withdrawal.